One of the advantages of using tree descriptions instead of trees in a grammar is the possiblility of underspecification. In a TDG for certain ambiguities there can be underspecified descriptions in the style of the underspecified representations Muskens derives in Muskens (1995) for quantifier scope ambiguities.
In principle the idea of TDGs is the following: The grammar consists of a set of tree descriptions, i.e. formulas in some first order logic, and the grammar is derivation based. Every derivation step is a conjunction of two formulas and additional equivalences of node names.
The formalism of Tree Description Grammars was introduced in Kallmeyer (1996). The main focus of the present paper will be on the formal properties of TDGs. I will show the decidability of lexicalized TDGs and the fact that the languages of TDGs are a true superset of tree adjoining languages.